Large language model (LLM)

A large language model is a piece of AI software trained on enormous amounts of text until it becomes good at predicting language. That one skill turns out to cover a lot: answering questions, summarizing documents, drafting emails, translating, and extracting information from messy text. ChatGPT, Claude, and Gemini are all products built on large language models.

What an LLM doesn't do matters just as much. It doesn't know your business unless you give it your information, it doesn't check facts unless it's connected to a source of facts, and it can state wrong things confidently (see hallucination). Most useful business AI is an LLM plus plumbing: your documents, your rules, and guardrails around what it's allowed to say.

Practical example: an accounting firm we'd typically work with doesn't need 'an LLM.' It needs the LLM connected to its own engagement letters and rate sheets so the model drafts client replies that are actually correct for that firm. The model is a commodity; the connection to your business is where the value lives.